Time-frequency system integrity monitoring method and device based on robust Kalman filter

Abstract

The present invention relates to a method and device for monitoring the integrity of a time-frequency system based on a robust Kalman filter. The method comprises: constructing a robust Kalman filter model. The state of the time difference information in the time-frequency system is estimated by the robust Kalman filter model, wherein the state comprises: time difference prediction deviation, frequency deviation and temperature variation coefficient. An observation model is constructed according to the state. The time difference prediction deviation and the frequency deviation are respectively outputted by the observation model to output the residual vectors corresponding to the time difference prediction deviation and the frequency deviation, and then the integrity is monitored. The time-frequency system generates an alarm prompt according to the residual vector and a preset fault detection scheme. The method can effectively improve the detection sensitivity and alarm efficiency of time-frequency system faults.

Description

Technical Field

[0001] The present invention relates to the technical field of time-frequency system integrity monitoring, and in particular to a time-frequency system integrity monitoring method and device based on a robust Kalman filter. Background Art

[0002] The health of the time and frequency system (hereinafter referred to as the time and frequency system) directly determines the performance of the navigation, positioning and timing services of the Global Navigation Satellite System (GNSS). The robustness and reliability of the time and frequency system are very important links to ensure the long-term stable and uninterrupted service of the satellite navigation system. Integrity monitoring of the time and frequency system can effectively improve the robustness and reliability of the time and frequency system.

[0003] At present, the development and research of integrity monitoring are mainly concentrated in the field of GNSS integrity monitoring. Receiver Autonomous Integrity Monitoring (RAIM) is one of the core technologies for realizing GNSS integrity monitoring. RAIM is a method for GNSS receivers to autonomously detect and eliminate errors based on redundant GNSS information based on consistency check theory. Commonly used RAIM algorithms include pseudorange comparison method, least squares residual method and parity vector method. The above three methods have good and equivalent effects in single fault scenarios. The relatively simple parity vector method is recommended by RTCASC-159 as the basic algorithm. At the same time, parameters such as maximum precision factor change, approximate radial error protection (ARP), and protection level (PL) are introduced to ensure the availability of RAIM. At the same time, based on the timing integrity requirements of timing receivers, a timing receiver autonomous integrity monitoring (T-RAIM) algorithm based on the least squares residual method is proposed.

[0004] The Kalman filter is an efficient recursive filter that is widely used in various fields, including power systems, communications, navigation, and medical imaging. It is based on a linear state space representation and processes noisy input and observation signals to obtain the system state or true signal. In practical applications, the Kalman filter has been proven to provide effective performance in a variety of complex environments. For example, in power systems, the Kalman filter is used for frequency estimation and error compensation of power metering devices; in the field of communications, it is used for carrier recovery and phase noise estimation; in navigation systems, the Kalman filter is a key component of the GNSS integrated inertial navigation system. With the development of technology, the Kalman filter has also derived a variety of improved versions. The Extended Kalman Filter (EKF) is able to process nonlinear system data. The robust Kalman filter aims to improve the filtering performance in the presence of abnormal observation errors and dynamic model errors.

[0005] At present, there are few methods for integrity monitoring of time-frequency systems, and there are only a few scattered methods for integrity monitoring of the time-frequency signals output by the time-frequency system. Existing research only focuses on the integrity monitoring of the time-frequency signals output by the time-frequency system. The integrity monitoring of the traditional time-frequency system as a whole is based on the consistency test theory, which only monitors the time difference information and has low monitoring sensitivity. Summary of the invention

[0006] Based on this, it is necessary to provide a time-frequency system integrity monitoring method and device based on a robust Kalman filter that can improve the alarm efficiency and accuracy of the time-frequency system in response to the above technical problems.

[0007] A time-frequency system integrity monitoring method based on a robust Kalman filter, the method comprising:

[0008] Construct a robust Kalman filter model.

[0009] The state of the time difference information in the time-frequency system is estimated by a robust Kalman filter model, wherein the state includes: time difference prediction deviation, frequency deviation and temperature variation coefficient.

[0010] Construct an observation model based on the state.

[0011] The time difference prediction deviation and frequency deviation are respectively output by the observation model to monitor the integrity of the residual vector corresponding to the time difference prediction deviation and the frequency deviation. The time-frequency system generates an alarm prompt based on the residual vector and the preset fault detection scheme.

[0012] A time-frequency system integrity monitoring device based on a robust Kalman filter, the device comprising:

[0013] Model building module, used to build a robust Kalman filter model.

[0014] The state acquisition module is used to estimate the state of the time difference information in the time-frequency system through the robust Kalman filter model, wherein the state includes: time difference prediction deviation, frequency deviation and temperature variation coefficient.

[0015] The observation model building module is used to build the observation model according to the state.

[0016] The integrity detection module is used for integrity monitoring of the time difference prediction deviation and the frequency deviation after the residual vectors corresponding to the time difference prediction deviation and the frequency deviation are output through the observation model respectively. The time-frequency system generates an alarm prompt based on the residual vector and the preset fault detection scheme.

[0017] The above-mentioned time-frequency system integrity monitoring method and device based on the robust Kalman filter firstly fully considers the influence of random noise, abnormal data points and temperature changes that may exist in the time-frequency system when constructing the robust Kalman filter model. Although the traditional Kalman filter has good state estimation ability, it has limited processing ability for abnormal data points (such as mutation signals or noise interference), which easily leads to distortion of the filtering result. Then, by introducing robustness, the weight of the observation noise can be dynamically adjusted during the filtering process, thereby suppressing the influence of abnormal data points, significantly improving the adaptability to complex dynamic environments and robustness to faults. In terms of obtaining the time difference information state of the time-frequency system, the robust Kalman filter estimates the key parameters of the system in real time, including time difference, frequency deviation and temperature variation coefficient. The time difference reflects the time error existing in the system during the synchronization process, and the frequency deviation characterizes the stability and accuracy of frequency synchronization. These state parameters are directly related to the overall performance of the time-frequency system. Secondly, through high-precision state estimation, a reliable foundation is laid for subsequent fault detection. Then, an observation model is constructed to further enhance the sensitivity and accuracy of monitoring. Through the improved observation model, the state estimation results of the robust Kalman filter are dynamically compared with the actual observation values ​​to accurately capture potential abnormal signal changes. The observation model can not only output error information in real time, but also has strong adaptability. It can adjust the detection parameters according to the different operating environments of the system to ensure that high detection performance can be maintained in different scenarios. Finally, in the fault detection and alarm prompt link, the source of the abnormal signal is quickly located by analyzing the observation error and the preset fault detection threshold. For complex fault types such as phase jump and frequency jump, this method can achieve efficient detection and accurate identification, avoiding the problems of missed detection and false detection caused by insufficient sensitivity in traditional methods. By combining with the preset alarm scheme, the system can send an alarm prompt to the user in real time when a fault occurs, ensuring that the user can understand the system status and take necessary measures in the first time.

[0018] In summary, this method forms a complete and efficient time-frequency system integrity monitoring process from model construction, state estimation, fault detection to alarm prompts. It not only improves the system's detection sensitivity and alarm efficiency for complex faults, but also significantly enhances the accuracy of fault identification, further ensuring the reliability and stable operation capability of the time-frequency system. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 1 is a flow chart of a method for monitoring integrity of a time-frequency system based on a robust Kalman filter in one embodiment;

[0020] Figure 2 A flowchart of time-frequency system integrity monitoring fusion assessment in one embodiment;

[0021] Figure 3 The figure is a structural block diagram of a time-frequency system integrity monitoring device based on a robust Kalman filter in one embodiment. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0023] In one embodiment, Figure 1 As shown, a time-frequency system integrity monitoring method based on a robust Kalman filter is provided, comprising the following steps:

[0024] Step 102: construct a robust Kalman filter model.

[0025] Step 104, estimating the state of the time difference information in the time-frequency system by using the robust Kalman filter model.

[0026] Among them, the status includes: time difference forecast deviation, frequency deviation and temperature variation coefficient.

[0027] Step 106, constructing an observation model according to the state.

[0028] Step 108, the time difference prediction deviation and the frequency deviation are respectively outputted by the observation model and then the residual vectors corresponding to the time difference prediction deviation and the frequency deviation are monitored for integrity, and the time-frequency system generates an alarm prompt according to the residual vector and a preset fault detection scheme.

[0029] In the above-mentioned time-frequency system integrity monitoring method based on the robust Kalman filter, first of all, when constructing the robust Kalman filter model, the influence of random noise, abnormal data points and temperature changes that may exist in the time-frequency system is fully considered. Although the traditional Kalman filter has good state estimation capabilities, its processing ability for abnormal data points (such as mutation signals or noise interference) is limited, which easily leads to distortion of the filtering results. Then, by introducing robustness, the weight of the observation noise can be dynamically adjusted during the filtering process, thereby suppressing the influence of abnormal data points, significantly improving the adaptability to complex dynamic environments and robustness to faults. In terms of obtaining the time difference information state of the time-frequency system, the robust Kalman filter estimates the key parameters of the system in real time, including time difference, frequency deviation and temperature variation coefficient. The time difference reflects the time error existing in the system during the synchronization process, and the frequency deviation characterizes the stability and accuracy of frequency synchronization. These state parameters are directly related to the overall performance of the time-frequency system. Secondly, through high-precision state estimation, a reliable foundation is laid for subsequent fault detection. Then, an observation model is constructed to further enhance the sensitivity and accuracy of monitoring. Through the improved observation model, the state estimation results of the robust Kalman filter are dynamically compared with the actual observation values ​​to accurately capture potential abnormal signal changes. The observation model can not only output error information in real time, but also has strong adaptability. It can adjust the detection parameters according to the different operating environments of the system to ensure that high detection performance can be maintained in different scenarios. Finally, in the fault detection and alarm prompt link, the source of the abnormal signal is quickly located by analyzing the observation error and the preset fault detection threshold. For complex fault types such as phase jump and frequency jump, this method can achieve efficient detection and accurate identification, avoiding the problems of missed detection and false detection caused by insufficient sensitivity in traditional methods. By combining with the preset alarm scheme, the system can send an alarm prompt to the user in real time when a fault occurs, ensuring that the user can understand the system status and take necessary measures in the first time.

[0030] In summary, this method forms a complete and efficient time-frequency system integrity monitoring process from model construction, state estimation, fault detection to alarm prompts. It not only improves the system's detection sensitivity and alarm efficiency for complex faults, but also significantly enhances the accuracy of fault identification, further ensuring the reliability and stable operation capability of the time-frequency system.

[0031] In one embodiment, the a priori temperature variation coefficient is calculated according to the combined model of the temperature compensation frequency deviation, and the time difference between the time-frequency links is corrected to obtain the a priori temperature variation coefficient and the corrected time difference:

[0032] ;

[0033] in, is the combined frequency deviation, For time, is the combined noise of any two time-frequency links, and its standard deviation is , To correct the time difference.

[0034] It is worth noting that the corrected time difference result is obtained to calculate the noise method and matrix Q, and the a priori temperature variation coefficient is obtained to construct the observation model. The a priori knowledge is used for secondary temperature control, so that the signal transmitted in the time-frequency link can be better de-temperatured and the signal's ability to resist temperature changes is improved. The noise standard deviation of the link is obtained to calculate the observation covariance R.

[0035] In one embodiment, the state equation of a single time-frequency source is defined as:

[0036] ;

[0037] ;

[0038] in, Indicates a moment in time. for The state variables of the time-frequency source transmitted by the time-frequency link at time instant, is the state transfer matrix, is the driving noise. The noise variance is calculated using the variance inversion algorithm according to the corrected time difference and the data sampling interval of the current time-frequency link, and the covariance matrix of the driving noise is obtained:

[0039] ;

[0040] in, is the data sampling interval, is the frequency white noise variance, is the frequency random walk noise variance, is the variance of the temperature coefficient, and Q is the covariance matrix of the driving noise. Define the observation equation for a single time-frequency source:

[0041] ;

[0042] ;

[0043] in, for Temperature change value at time, for The observation matrix at the moment; the robust Kalman filter model is constructed based on the state equation, the observation equation of a single time-frequency source, and the robust equivalent weight function:

[0044] ;

[0045] in, and They are Moment and The estimated state value of the time-frequency source of the time-frequency link at time, for Time has come The predicted value of the state at time for The observed quantity of the time-frequency source of the time-frequency link at time, for The observation matrix at time, is the robust observation vector covariance, for Time has come The a priori estimated covariance of time, and They are Moment and The posterior estimated covariance at time , is the robust Kalman gain.

[0046] It is worth noting that the temperature change value is the result of subtracting the reference temperature from the measurement result of the high-resolution temperature measurement module, and the reference temperature is generally 25°C.

[0047] In one embodiment, the expansion factor of the time-frequency link is constructed using a robust equivalent weight function:

[0048] ;

[0049] in, is the standardized residual, For the The residual of the observation vector, For the The noise standard deviation of the observation vector. The covariance of the observation vector after epoch anti-error is obtained according to the expansion coefficient factor:

[0050] ;

[0051] in, is the observation vector covariance, is the expansion coefficient factor, For the i The standard deviation of the noise of the link.

[0052] In one embodiment, the estimation results of the time difference and the frequency deviation are used as observation quantities to construct an observation model:

[0053] ;

[0054] in, For the observed dimensional vector, for dimensional coefficient matrix, is the 1-dimensional observation result, for dimensional observation noise vector, is the number of observations.

[0055] In one embodiment, the time difference forecast deviation and the frequency deviation are respectively input into the observation model, and the residual vector of the observation quantity is obtained by using the least square method:

[0056] ;

[0057] Where v is the residual vector of the time difference prediction deviation or frequency deviation, and I is The unit matrix of dimension H is The coefficient matrix of dimension W is dimensional observation weight matrix, for The time difference prediction deviation and frequency deviation are monitored for integrity according to the residual vector of the observed quantity, and the time difference monitoring results and frequency difference monitoring results are output.

[0058] In one embodiment, the time-frequency system uses an AND gate method to fuse the time difference monitoring results and frequency difference monitoring results obtained by the residual vector integrity monitoring to obtain a test statistic, and generates an alarm prompt based on the test statistic and a preset fault detection scheme. The fault detection scheme includes: fault detection and fault identification based on the residual sum of squares. Fault detection is:

[0059] ;

[0060] in, is the equivalent observation error, is the fault detection threshold, is the preset false alarm probability; if SSE / If it is greater than the fault detection threshold, it is judged that the observation corresponding to the current weight residual is inconsistent with the detection, and a fault detection alarm is generated. Fault identification is:

[0061] ;

[0062] in, is the test statistic, T d is the fault identification threshold; if the test statistic is greater than the fault identification threshold, it is judged that the observation corresponding to the current test statistic is faulty and a fault alarm is generated.

[0063] It is worth noting that the post-test unit weighted mean error Calculated by the residual sum of squares (Sum of Square Error, SSE):

[0064] ;

[0065] If the components of the observation noise vector ε are independent normally distributed random errors with a mean of 0 and a variance of , according to statistical distribution theory, we can make a binary hypothesis:

[0066] H 0 (no fault): ,but .

[0067] H 1 (Faulty): ,but .

[0068] Given the false alarm probability P FA , the following probability equation holds:

[0069] ;

[0070] The above formula can be used to determine SSE / The detection threshold T , thus determining The detection threshold . The real-time calculation and Compare, if , it means that an inconsistency is detected and a warning is issued to the user.

[0071] In addition, based on the relationship between residual and observation error, the constructed statistics are as follows:

[0072] ;

[0073] In the formula, . for No. i Line i The value of the column. d i Make a binary hypothesis:

[0074] H 0 (no fault): ,but .

[0075] H 1 (Faulty): ,but .

[0076] in, δ i is the statistical offset parameter, if i The time difference deviation of each observation is b i ,but

[0077] ;

[0078] In the formula, For W i Line i The value of the column. n The observations can be obtained n test statistic, given the overall false alarm probability P FA , the false alarm probability of each statistic is P FA / n . Thus, the following probability equation holds:

[0079] ;

[0080] The detection limit can be calculated by the above formula: T d For each test statistic d i , respectively T d Compare, if d i > T d , indicating that i An observation has a fault and the user is notified that it needs to be corrected.

[0081] In one embodiment, if Figure 2 As shown, a fusion evaluation method for integrity monitoring of a time-frequency system is provided. The specific steps are as follows: using the time difference measurement results, the time difference and frequency deviation are estimated, and the time difference prediction deviation is calculated. When the time-frequency system is fault-free, the time difference prediction deviation is consistent with the measurement noise. The time difference prediction deviation can be used as a residual for gross error detection to achieve integrity monitoring of the time difference. At the same time, the consistency test theory is used to achieve integrity monitoring of the frequency deviation. The time difference integrity monitoring results and the frequency deviation integrity monitoring results are fused and evaluated to achieve integrity monitoring of the time-frequency system.

[0082] When the number of observations is less than 2, there is no redundant information and the integrity monitoring algorithm is not applicable. When the number of observations is not less than 2, the error protection level parameter is used to determine whether the integrity monitoring algorithm is applicable.

[0083] Assume i There are faults in observations, and their deviation is b i , then the test statistic SSE / Obey the non-center x 2 Distribution, ignoring the influence of normal errors, its non-central parameter is expressed as:

[0084] ;

[0085] Multiply the numerator and denominator of the above formula by ,but

[0086] ;

[0087] In the formula, , A i Given the probability of missed detection P MD , the probability equation should be satisfied to obtain the non-centralized parameter λ;

[0088] ;

[0089] make is the characteristic slope line for each observation, which gives the relationship between the error and the detection statistic. For a given systematic measurement error, the observation with the larger slope has the smaller detection statistic and is the most difficult to detect. SLOPE max is the slope of the observation that is most difficult to detect.

[0090] Therefore, the non-centrality parameter can be changed to

[0091] ;

[0092] Pick The maximum value of , and then take the square root of both sides of the equation to get

[0093] ;

[0094] Right now

[0095] ;

[0096] Will Determined as a protection level (PL), then

[0097] ;

[0098] The protection level is actually the maximum value of the system error vector. When this maximum value is less than the protection threshold (AL) set by the user, the integrity monitoring algorithm is available.

[0099] It should be understood that although Figure 1-Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 1-Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0100] In one embodiment, Figure 3 As shown, a time-frequency system integrity monitoring device based on a robust Kalman filter is provided, comprising: a model building module 302, a state acquisition module 304, an observation model building module 306 and an integrity detection module 308, wherein:

[0101] The model building module 302 is used to build a robust Kalman filter model.

[0102] The state acquisition module 304 is used to estimate the state of the time difference information in the time-frequency system through the robust Kalman filter model, wherein the state includes: time difference prediction deviation, frequency deviation and temperature variation coefficient.

[0103] The observation model building module 306 is used to build an observation model according to the state.

[0104] The integrity detection module 308 is used for integrity monitoring of the time difference prediction deviation and the frequency deviation after the residual vectors corresponding to the time difference prediction deviation and the frequency deviation are output by the observation model respectively. The time-frequency system generates an alarm prompt according to the residual vector and the preset fault detection scheme.

[0105] For the specific limitations of the time-frequency system integrity monitoring device based on the robust Kalman filter, please refer to the limitations of the time-frequency system integrity monitoring method based on the robust Kalman filter in the above text, which will not be repeated here. Each module in the above-mentioned time-frequency system integrity monitoring device based on the robust Kalman filter can be implemented in whole or in part through software, hardware and a combination thereof. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0106] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0107] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0108] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0109] The above-mentioned embodiments only express several implementation modes of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the invention. It should be pointed out that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the attached claims.

Claims

1. A time-frequency system integrity monitoring method based on a robust Kalman filter, characterized in that: The method comprises: Construct a robust Kalman filter model; The state of the time difference information in the time-frequency system is estimated by the robust Kalman filter model, wherein the state includes: time difference prediction deviation, frequency deviation and temperature variation coefficient; constructing an observation model based on the state; The time difference prediction deviation and the frequency deviation are respectively outputted by the observation model as residual vectors corresponding to the time difference prediction deviation and the frequency deviation for integrity monitoring, and the time-frequency system generates an alarm prompt according to the residual vector and a preset fault detection scheme.

2. The method according to claim 1, characterized in that Before the step of building the robust Kalman filter model, it also includes: The a priori temperature variation coefficient is calculated based on the combined model of temperature compensation frequency deviation, and the calculation formula is as follows: In the formula, Represents the difference in delay between two links of the time-frequency signal, represents the combined frequency deviation, t Indicates time, A represents the combined temperature variation coefficient, represents the change in ambient temperature of the time-frequency system, represents the combined noise of the two links; Combined temperature coefficient of variation A As the a priori temperature variation coefficient and correcting the time difference between the time-frequency links, the corrected time difference is obtained: in, is the combined frequency deviation, For time, is the combined noise of any two time-frequency links, To correct the time difference.

3. The method according to claim 2, characterized in that Construct a robust Kalman filter model, including: Define the state equation for a single time-frequency source: in, Indicates a moment in time. for The state variables of the time-frequency source transmitted by the time-frequency link at time instant, is the state transfer matrix, To drive the noise; The noise variance is calculated using the variance inversion algorithm according to the corrected time difference and the data sampling interval of the current time-frequency link to obtain the covariance matrix of the driving noise: in, is the data sampling interval, is the frequency white noise variance, is the frequency random walk noise variance, is the variance of temperature coefficient variation, Q is the covariance matrix of driving noise; Define the observation equation for a single time-frequency source: in, for Temperature change value at time, for The observation matrix at time; A robust Kalman filter model is constructed according to the state equation, the observation equation of the single time-frequency source, and the robust equivalent weight function: in, and They are Moment and The estimated state value of the time-frequency source of the time-frequency link at time, for Time has come The predicted value of the state at time for The observed quantity of the time-frequency source of the time-frequency link at time, for The observation matrix at time, is the robust observation vector covariance, for Time has come The prior estimated covariance of time, and They are Moment and The posterior estimated covariance at time , is the robust Kalman gain.

4. The method according to any one of claims 1 to 3, characterized in that: The expansion coefficient factor of the time-frequency link is constructed using the robust equivalent weight function: in, is the standardized residual, For the The residual of the observation vector, For the The standard deviation of the noise of the observation vector; The covariance of the observation vector after epoch robustification is obtained according to the expansion coefficient factor: in, is the observation vector covariance, is the expansion coefficient factor, For the i The standard deviation of the noise of the link.

5. The method according to claim 4, characterized in that An observation model is constructed according to the state, including: The estimation results of the time difference and the frequency deviation are used as observation quantities to construct an observation model: in, For the observed dimensional vector, for dimensional coefficient matrix, is the 1-dimensional observation result, for dimensional observation noise vector, is the number of observations.

6. The method according to claim 5, characterized in that The time difference prediction deviation and the frequency deviation are respectively subjected to integrity detection after the residual vectors corresponding to the time difference prediction deviation and the frequency deviation are outputted by the observation model, including: The time difference forecast deviation and the frequency deviation are respectively input into the observation model, and the residual vector of the observation quantity is obtained by using the least square method: Wherein, v is the residual vector of the time difference prediction deviation or frequency deviation, and I is The unit matrix of dimension H is The coefficient matrix of dimension W is dimensional observation weight matrix, for dimensional observation noise vector; The time difference prediction deviation and the frequency deviation are respectively monitored for integrity according to the residual vector of the observed quantity, and a time difference monitoring result and a frequency difference monitoring result are output.

7. The method according to claim 6, characterized in that The time-frequency system generates an alarm prompt according to the residual vector and a preset fault detection scheme, including: The time-frequency system fuses the time difference monitoring result and the frequency difference monitoring result obtained by the residual vector integrity monitoring in an AND gate manner to obtain a test statistic, and generates an alarm prompt according to the test statistic and a preset fault detection scheme; The fault detection scheme includes: fault detection and fault identification based on residual sum of squares; The fault detection is: in, is the equivalent observation error, is the fault detection threshold, is the preset false alarm probability; if SSE / If the error is greater than the fault detection threshold, it is determined that the observed value corresponding to the current weight residual is inconsistent with the detection, and a fault detection alarm is generated; The fault identification is: in, d is the test statistic, T d is the fault identification threshold; if the test statistic is greater than the fault identification threshold, it is determined that the observation corresponding to the current test statistic is faulty, and a fault alarm is generated.

8. A time-frequency system integrity monitoring device based on a robust Kalman filter, characterized in that: The device comprises: A model building module, used to build a robust Kalman filter model; A state acquisition module, used for estimating the state of the time difference information in the time-frequency system through the robust Kalman filter model, wherein the state includes: time difference prediction deviation, frequency deviation and temperature variation coefficient; An observation model building module, used to build an observation model according to the state; The integrity detection module is used for integrity monitoring of the time difference prediction deviation and the frequency deviation after the residual vectors corresponding to the time difference prediction deviation and the frequency deviation are output by the observation model respectively, and the time-frequency system generates an alarm prompt according to the residual vector and a preset fault detection scheme.

Images